Telescope Array search for EeV photons Yana Zhezher (on behalf of - - PowerPoint PPT Presentation
Telescope Array search for EeV photons Yana Zhezher (on behalf of Mikhail Kuznetsov), Oleg Kalashev and Grigory Rubtsov for the Telescope Array collaboration ICRR, Univ. of Tokyo & INR RAS TAUP-2019 11 September 2019 Supported by Russian
for the Telescope Array collaboration ICRR, Univ. of Tokyo & INR RAS
Supported by Russian Science Foundation
Search for photons with TA SD 1
based on Abbasi et al., arXiv:1904.00300
Stokes et al, Astropart.Phys.35:759,2012
Search for photons with TA SD 3
5 10 15 20 25 30 35 40 1 2 3 4 5 6 Time, µs γ 20080514 32259.903427 (17.13, 23.44) ndet=9 angles: 50.6, 128.6 S800=12.9 upper lower
5 10 15 20 25 30 1 2 3 4 5 6 Time, µs γ 20080607 120538.696262 (21.72, 7.06) ndet=9 angles: 47.1, 93.5 S800=13.6 upper lower
Photon-induced showers: ◮ arrive younger ◮ contain less muons ◮ multiple SD observables affected: front curvature, Area-over-peak, χ2/d.o.f., etc.
Pierre Auger Collaboration, Phys.Rev.Lett. 100 (2008) 211101
Ros, Supanitsky, Medina-Tanco et al. Astropart.Phys. 47 (2013) 10
Search for photons with TA SD 5
◮ The Boosted Decision Trees (BDT) technique is used to build p-γ classifier based on multiple observables.
Telescope Array, Astropart. Phys. 110, 8 (2019); PRD 99, 022002 (2019)
◮ root::TMVA is used as a stable implementation.
PoS ACAT 040 (2007), arXiv:physics/0703039
◮ BDT is trained with Monte-Carlo sets: γ (signal) and p (background)* ◮ BDT classifier is used to convert the set of observables of each event to a number ξ ∈ [−1 : 1] ◮ ξ is available for one-dimensional analysis.
* MC set is split into 3 equal parts: (I) for training the classifier, (II) for ξ-cut
ξ
0.2 0.4 0.6 0.8 1
10 1 10
2
10
3
10
4
10
/eV)>18.0
γ
log(E
gamma MC proton MC data
/eV)>18.0
γ
log(E
ξ
0.2 0.4 0.6 0.8 1
10 1 10
2
10
3
10
4
10
/eV)>18.5
γ
log(E
gamma MC proton MC data
/eV)>18.5
γ
log(E
ξ
0.2 0.4 0.6 0.8 1
10 1 10
2
10
3
10
/eV)>19.0
γ
log(E
gamma MC proton MC data
/eV)>19.0
γ
log(E
ξ
0.2 0.4 0.6 0.8 1
10 1 10
2
10
/eV)>19.5
γ
log(E
gamma MC proton MC data
/eV)>19.5
γ
log(E
Diffuse photon search with the TA SD: [TA], Astropart.Phys. 110 (2019) 8-14
The way to improve the photon search sensitivity:
Hadron background is highly isotropic ⇓ Assume that photons are emitted by point source ⇓ In angular vicinity of the source the photon/hadron ratio would be larger than in full TA field of view ⇓
Specific photon source hypotheses could be tested by search in certain directions stacked ◮ Dwarf spheroidal galaxies (heavy DM decay hypothesis)
Search for photons with TA SD 8
Diffuse photon search with the TA SD: [TA], Astropart.Phys. 110 (2019) 8-14
The way to improve the photon search sensitivity:
Hadron background is highly isotropic ⇓ Assume that photons are emitted by point source ⇓ In angular vicinity of the source the photon/hadron ratio would be larger than in full TA field of view ⇓
Specific photon source hypotheses could be tested by search in certain directions stacked ◮ Dwarf spheroidal galaxies (heavy DM decay hypothesis)
Search for photons with TA SD 9
◮ Independent search for γ in each skymap direction ◮ The angular size of the each search region is equal to the γ angular resolution:
Eγ ≥, eV 1018.0 1018.5 1019.0 1019.5 1020.0 ang.res. 3.00◦ 2.92◦ 2.64◦ 2.21◦ 2.06◦
◮ The skymap is pixelized into 12288 directions with HEALPix (7868 in TA field of view) Optimisation of MVA-cut for γ flux upper-limit: ◮ Assume the flux consists of protons only (null hypothesis): Ftotal = Fp ◮ Optimize the ξ-cut separately for the best upper-limit in each direction using MC p and MC γ ◮ E−2 γ-spectrum is assumed
Search for photons with TA SD 10
Photon flux upper-limit, E > 1 EeV
0.242201
km
2yr 1
90° 180° 270° +90°
0°
Eγ ≥, eV Fγ ≤, km−2yr−1 1018.0 0.094 1018.5 0.029 1019.0 0.010 1019.5 0.0071 1020.0 0.0058 Pierre Auger: Fγ ≤ 0.035 km−2yr−1 (1◦ ang.res., 1017.3 ≤ E ≤ 1018.5 eV)
Search for photons with TA SD 11
Photon flux upper-limit, E > 3.16 EeV
0.071541
km
2yr 1
0° 90° 180° 270° +90°
Photon flux upper-limit, E > 10 EeV
0.029244
km
2yr 1
0° 90° 180° 270° +90°
Photon flux upper-limit, E > 31.6 EeV
0.01895
km
2yr 1
0° 90° 180° 270° +90°
Photon flux upper-limit, E > 100 EeV
0.01406
km
2yr 1
0° 90° 180° 270° +90°
The results for all points are available in table form with paper
Abbasi et al., arXiv:1904.00300
It can be used to constrain models of UHECR sources.
Excess significance, E > 1 EeV
2.722
0° 90° 180° 270° +90°
Eγ ≥, eV
1018.0 2.72 σ 1018.5 2.71 σ 1019.0 2.89 σ 1019.5 2.76 σ 1020.0 3.43 σ
The excesses are insignificant, given the large number of trials, no evidence for the photon signal on the current level of sensitivity.
Excess significance, E > 3.16 EeV
2.707
0° 90° 180° 270° +90°
Excess significance, E > 10 EeV
2.89
0° 90° 180° 270° +90°
Excess significance, E > 31.6 EeV
2.759
0° 90° 180° 270° +90°
Excess significance, E > 100 EeV
3.425
0° 90° 180° 270° +90°
The largest pre-trial excess is 3.43 σ for Eγ ≥ 1020.0 eV located at {α = 155.3◦, δ = 60.4◦}
Probe for the possible decay of heavy dark matter (HDM) ◮ HDM decay produce significant amount of photons in any model
◮ DM is abundant in dwarf galaxies (Galactic Center is outside the TA field of view) ◮ Target source set: 21 dwarf galaxies — satellites of Milky Way
◮ Search for γ in stacked skymap pixels of dwarf galaxies (pixel size = γ ang.res.) Results No evidence for photon signal (Ncand.
γ
= 0 at all energies)
Eγ, eV 1018.0 1018.5 1019.0 1019.5 1020.0 F γ
UL, km−2yr−1
0.15 0.057 0.014 0.0076 0.0052
These results can be used to constrain HDM models
Search for photons with TA SD 15
Probe for the possible decay of heavy dark matter (HDM) ◮ HDM decay produce significant amount of photons in any model
◮ DM is abundant in dwarf galaxies (Galactic Center is outside the TA field of view) ◮ Target source set: 21 dwarf galaxies — satellites of Milky Way
◮ Search for γ in stacked skymap pixels of dwarf galaxies (pixel size = γ ang.res.) Results No evidence for photon signal (Ncand.
γ
= 0 at all energies)
Eγ, eV 1018.0 1018.5 1019.0 1019.5 1020.0 F γ
UL, km−2yr−1
0.15 0.057 0.014 0.0076 0.0052
These results can be used to constrain HDM models
Search for photons with TA SD 16
Improve γ-search sensitivity with neural networks: PoS(ICRC2019)304 Increase γ-search exposure with TAx4
Search for photons with TA SD 17
◮ The directional upper-limits for photons with E > 1018 eV are set ◮ The directional upper-limits for E > 1018.5 eV are set for the first time ◮ No significant evidence for photon signal was found in blind search ◮ The results can be used to constrain UHECR sources models
◮ The photons from stacked dwarf galaxies were searched — no candidates were found ◮ The flux upper-limits for dwarf galaxies can be used to constrain heavy dark matter models
Search for photons with TA SD 18
◮ The directional upper-limits for photons with E > 1018 eV are set ◮ The directional upper-limits for E > 1018.5 eV are set for the first time ◮ No significant evidence for photon signal was found in blind search ◮ The results can be used to constrain UHECR sources models
◮ The photons from stacked dwarf galaxies were searched — no candidates were found ◮ The flux upper-limits for dwarf galaxies can be used to constrain heavy dark matter models
Search for photons with TA SD 19
Search for photons with TA SD 20
The geometry reconstruction of events is crucial for point source search. There is a bias in zenith angle θ reconstruction. We correct all data and MC events for the mean value of this bias: θrec. − θtrue Angular reconstruction for photon primaries (used in this search) Eγ, eV θrec. − θtrue
1018.0 −2.25◦ 3.00◦ 1018.5 −2.24◦ 2.92◦ 1019.0 −2.16◦ 2.64◦ 1019.5 −2.06◦ 2.21◦ 1020.0 −1.72◦ 2.06◦
Search for photons with TA SD 21
◮ Independent search for γ in each skymap direction ◮ The angular size of the search region is equal to the γ angular resolution ◮ The skymap is pixelized into 12288 directions with HEALPix (7868 in TA field of view) Optimisation of MVA-cut for γ flux upper-limit: ◮ Assume the flux consists of protons only (null hypothesis): Ftotal = Fp ◮ Optimize the ξ-cut separately in each constant declination band δi ± ang.res. using MC p and MC γ ◮ The optimization is for the minimum upper-limit: F UL
γ
=
µUL
Poisson(Nξ p )
Aeff
γ
,
where Nξ
p = {number of MC p events passing the ξ-cut};
µUL
Poisson = {upper-limit of the Poisson mean}; Aeff γ = {Effective exposure of the experiment to γ for the
given values of event quality cuts and ξ-cut}
Search for photons with TA SD 22